Hailo-8L on Pi5 – Need SSD MobileNetV2 .so File but Cannot Compile

serein · November 29, 2025, 4:18am

Hello, I am working on deploying an SSD MobileNetV2 object detection model on a Raspberry Pi 5 + Hailo-8L for my research project. I have run into the following issues:

I do not have the .so file for SSD MobileNetV2.
I got an.so file from HailoRT for SSD MobileNet, but it’s for V1 and not V2. The compiled model of V2 cannot be used with this because V1.so looks for a specific tensor which exists in V1.hef and not in V2.hef. I’m kind of new to this so I do not know if there are any workarounds here to fix it. So, I’m basically looking for a .so file dedicated for V2. From the documentation, the .so file must be generated using the Hailo Dataflow Compiler, which leads me to number 2…
I cannot compile my own .so file.
The software suite requires atleast 16GB RAM, with 32GB recommended. Unfortunately, my laptop has exactly 16GB RAM and I am using VirtualboxUbuntu, and you know what happens from there.

My team and I really need to use SSD MobileNetV2 since that’s what’s in our approved research proposal. I just want to know if there is an official precompiled .so for V2 for the Hailo 8L device? If not, is there any alternative workflow for users who do not have a machine capable of running the AI Software Suite + Dataflow Compiler? Is there a cloud-based compilation option, or can someone from the community share advice on compiling when hardware resources are limited?

omria · November 30, 2025, 10:03am

Hey @serein,

Welcome to the Hailo Community!

What do you mean so ?

We got the compiled hef file for the hailo8l ssd_mobilenet_v2 right here:
https://hailo-model-zoo.s3.eu-west-2.amazonaws.com/ModelZoo/Compiled/v2.17.0/hailo8l/ssd_mobilenet_v2.hef

If you mean the postprocessing - that’s pretty straightforward. The standard hailortpp should do the trick, but if you run into any issues or want to customize it, you can check out this guide on writing your own postprocess:

github.com/hailocs/hailo-apps-infra

doc/developer_guide/writing_postprocess.md

dev

# Writing Your Own Post-Process for Hailo Apps

So you want to add a new network to Hailo-apps-infra that isn't supported yet? You've come to the right place! This guide will walk you through creating your own post-process. Don't worry—it's easier than it sounds.
This guide describes how to create a post-process function written in C++ that can be used in the HailoApps framework. This is done using the 'hailofilter' element.
**Note** For python post-process, you can add your post-process function as a python callback function.

## Getting Started
The post-process function should read the outputs of the HailoNet element, which are tensors attached to the HailoROI object. The post-process function should remove the tensors from the HailoROI object and replace them with the post-process results. i.e. HailoDetection, HailoClassification, etc.

### Where to Put Your Files

First things first - navigate to this directory:
```
hailo_apps_infra/hailo_cpp_postprocess/cpp
```

All your new post-processing magic will happen in the `postprocesses/` folder.

### Creating Your Header File

This file has been truncated. show original

Let me know if you need any help!